Free App Now Runs 15 More AI Models on Your Own Laptop

A popular free piece of software called llama.cpp released an update on September 18, labeled b11039. Think of llama.cpp as a universal media player for AI — like VLC, but for chatbot brains. Instead of sending your questions to a company's servers, it runs the AI directly on your own laptop, desktop, or phone. This update adds ready-made support for 15 more AI model families, so many more of them work out of the box instead of needing custom work.

The headline fix involves a technical phrase: sliding window attention. In plain terms, instead of remembering an entire long conversation at once, a model can focus on just the most recent chunk of text, which saves a lot of memory. Some model creators describe that chunk as one repeating pattern, while others list it out layer by layer. llama.cpp could only understand the first format and silently ignored the second, quietly applying wrong defaults. Now it reads both correctly, so converted models produce identical results to the originals.

The update also ships ready-to-install builds for nearly every kind of computer: Windows, Mac (both Intel and Apple Silicon), Linux, Android, and iPhone. There are extra speed options for Nvidia and AMD graphics cards, plus specialized versions for Intel chips and Chinese server hardware. It's mainly aimed at developers, researchers, and hobbyists who want private, offline AI that never phones home.

The catch is real: this is not a consumer product. There's no friendly installer or chat window. You generally work from a command line, download multi-gigabyte model files yourself, and troubleshoot graphics drivers when things break. The payoff, though, is genuine — no monthly subscription, nothing you type leaving your device, and it works with no internet connection at all.